Is AI Overhyped?

[Astronuc]

In The Street, AI is getting worse as Google and Anthropic nerf AI models and limit usage
Evidence is mounting that the subsidization of AI is ending in real time. Will businesses still shell out for it?
https://www.thestreet.com/latest-ne...-and-anthropic-nerf-ai-models-and-limit-usage

• Users report AI models from Anthropic, Google, and OpenAI are becoming worse or more restrictive.
• Companies introduce higher prices, usage limits, and downgrade product capabilities, sparking user frustration.
• Loss of initial subsidies and rising costs may jeopardize broad AI adoption and business viability.

The rise of artificial intelligence technology a la OpenAI, Anthropic, and Google has come by many names, but one of these is new: “worse.”

That is what users are saying about AI chatbots at a time where both the U.S. stock market is turbo bullish on the AI industry and fast-growing and unprofitable AI labs prepare for potentially blockbuster IPOs.

Anthropic, in many ways, brandished itself as a firm with a superior moral framework by standing up to the U.S. Government and setting limits on how its AI models could be used. Of course, much of that goodwill was completely burned by recent controversies with the company.

Anthropic had a billing controversy by charging customers per token for using popular third-party tools. Anthropic apparently downgraded reasoning effort in Claude and had a series of other issues which made their model 'dumber'. Did Anthropic address/resolve the issues when OpenAI introduced their GPT-5.5 model?

Google announced a series of new products, unveiled a controversial change to its search engine to make it ‘AI-first’, and touted a recent price reduction for its AI plans. But customers received an email on Wednesday with some bad news: your usage will be capped.

Effective immediately, Google announced that Gemini app users would face “compute-based usage limits.” In other words, if you ask Gemini too much, they’ll lock you out for five hours, assuming you haven’t somehow hit a fairly unknown weekly limit. One consolation for AI Pro subscribers is that they get a “4x higher usage limit than non-subscribers.” AI Ultra consumers get even more than that, but at a significantly higher price.

What about OpenAI and their products, e.g., ChatGPT? Will the make similar changes?

It seems as AI companies pursue profit, access will be governed by subscription fees. High end users, who can afford it, will get greater access than those who can only afford lower subscription rates. Maybe there will be 'preferred' customers.

 

[Hornbein]

The AI investors will want their trillion dollar investments back and plenty more on top of that. Once the customers are hooked prices go up and quality goes down.

 

[pinball1970]

I am not aware of this problem, I am aware of the eccentric Erdos.

This is a first for Ai?

https://arxiv.org/abs/2605.20579

Edit. A critique here https://cdn.openai.com/pdf/74c24085-19b0-4534-9c90-465b8e29ad73/unit-distance-remarks.pdf

 

[Greg Bernhardt]

This feels like a problem

 

[sbrothy]

AI is a tool, in my opinion. And just like every other tool, it has its advantages and disadvantages. It will be extremely helpful in areas like archeology or diagnostics, and of limited use in creativity, despite the fact that it can already mimic results in, for example, photography. Here is an article about it:

Artificial intelligence and the future of work: Will AI replace our jobs?



Is there anything they won't use for this purpose?



No. We are simply too many.

Also, we still have our finger on the power switch, The Matrix and Terminator notwithstanding.

 

[Matterwave]

AI is certainly over hyped, but even if it is significantly over hyped it can still be totally transformative for society. The hype is just that strong.

Reading statements from Dario/Sam etc. from last year at this time I would have expected all white collar jobs to be gone by now.

 

[PeroK]

AI is certainly over hyped, but even if it is significantly over hyped it can still be totally transformative for society. The hype is just that strong.

Reading statements from Dario/Sam etc. from last year at this time I would have expected all white collar jobs to be gone by now.

The sober, expert assessment is that AI could totally transform our civilization in 10-20 years. Certainly, almost all IT jobs could be gone in ten years. And perhaps the vast majority of white-collar jobs in twenty years.

 

[Stavros Kiri]

The sober, expert assessment is that AI could totally transform our civilization in 10-20 years. Certainly, almost all IT jobs could be gone in ten years. And perhaps the vast majority of white-collar jobs in twenty years.

Also, there's the [predicted] issue of "singularity point/event" [related to the appearance of ASI etc.] (some predictions around the same time frame ~ 2040-45 or -50s etc. ... [?], depending ...), ... if ..., if ...

 

[gleem]

The AI Futures Project is a nonprofit organization that studies the development of AI and attempts to predict the consequences of AI progress through its projected scenario AI 2027, which was published last year.

They assume two AI companies, which they call OpenBrain (US) and DeepCent (China), will predominate. If you read AI 2027, they have a large overall uncertainty for the time for the development of AGI, but they have the highest probability of AGI in the time from 2028 to 2030. In the video below, they look at the possible near-term development of AGI, something we can experience now, as they see 2026 as a period of rapid progress despite the prediction of current models hitting a brick wall. Last month, Anthropic's Mythos became the best agent for finding software bugs and security risks.

In the scenario, coding is important in the progress toward AGI. Currently, the two best coding agents are OpenAI's Codex (powered by GPT5.5) and Anthropic's Claude Code (powered by Claude Opus 4.8), both released earlier this year. Could OpenAI or Anthropic be AI Future's OpenBrain?

Some say not to worry because we have our finger on the power switch, but who is we, and when will we push it

https://www.youtube.com/watch?v=k_onqn68GHYthis

 

[Borg]

4 commercials within the first minute? I see someone's goal.

 

[Filip Larsen]

Some say not to worry because we have our finger on the power switch, but who is we, and when will we push it

Indeed. The argument that "the progression of AI development will remain safe because we can always stop developing and/or using it", with all the problematic parts of argument marked out, is obviously a fallacy.

The comparison between AI technology and nuclear technology in terms of potential safety consequences for human society is very apt, so one only need to ask if the same argument is valid if applied to the development history of nuclear technology while imagining that nuclear fuel itself could be produced solely by applying enough compute power.

 

[jack action]

The comparison between AI technology and nuclear technology in terms of potential safety consequences for human society is very apt, so one only need to ask if the same argument is valid if applied to the development history of nuclear technology while imagining that nuclear fuel itself could be produced solely by applying enough compute power.

Nuclear fueling itself might even have been easier than that, according to the experts and their knowledge at the time:

Three years before the Manhattan Project's first nuclear test, a small group of physicists considered the idea that a nuclear explosion set off in the air might set off a chain reaction igniting the atmosphere. The notion was first suggested by physicist Edward Teller, who spoke with colleagues including Hans Bethe and, yes, Oppenheimer.

Teller’s concern, roughly speaking, was that the energy from the blast might be sufficient to fuse nitrogen in the atmosphere into carbon and oxygen. The idea sounds silly today (those are the sorts of fusion reactions that occur inside the cores of massive late-stage stars), but at the time no one was entirely sure what would happen when the bomb went off.

Physicists crunched the numbers and were pretty certain the atmosphere wouldn’t ignite. The math just didn’t work out. In fact, before the test, Enrico Fermi casually offered a bet that a couple of scientists reportedly took. The attitude surrounding the possibility of fire in the sky was, apparently, casual.

Still, the possibility was taken seriously and investigated thoroughly. Arthur Compton, a Nobel Prize-winning physicists once said of the question, “It would be the ultimate catastrophe. Better to accept the slavery of the Nazis than to run the chance of drawing the final curtain on mankind!” He only agreed to the test after calculating the odds of a runaway reaction at less than one in 3 million, per Scientific American.

The experts - about something nobody ever had experience with - didn't know for sure it wouldn't have created human extinction, but they went with it anyway. And they didn't ask anyone for permission either; it was all out of the public eye. Imagine these experts sharing these concerns with the public before conducting their experiment, stating, "We're pretty certain the atmosphere wouldn't ignite."

So the comparison between nuclear and AI, while imagining that nuclear fuel itself is very valid.

But this was based on pure physics with math, not some doom's-day scenario - propagated through social media - involving politics, manipulation, and humans voluntarily submitting to machines, created by a vivid imagination.

 

[Filip Larsen]

The experts - about something nobody ever had experience with - didn't know for sure it wouldn't have created human extinction, but they went with it anyway. And they didn't ask anyone for permission either; it was all out of the public eye.

With the comparison I was not thinking of the hard singularity scenario. I am thinking about how various groups of people will surely misuse the technology to their own ends (including weaponizing) in disregard of long term safety.

But now you mention it the example you gave it is relevant. The point here is that at any given point in time there will surely always be someone willing to "take the risk" of advancing a potential dangerous technology that cannot be undone later. Right now we see AI technology being weaponized in many different ways with the intentional purpose and capability of bringing harm to other people and this capability is so very easy to copy for anyone seeking the same capability (quite unlike nuclear weapon). The potential for uncontrollable misuse of this technology is simply huge on a scale never seen before.

 

[Stavros Kiri]

The potential for uncontrollable misuse of this technology is simply huge on a scale never seen before.

... and especially with the new wars and [potentially (but perhaps not as of yet) uncontrollable] conflicts seen after 2022 up to today [(perhaps more than ever seen before after WW2 ... (?))] ... !

 

[gleem]

Pope Leo in his Encyclical "Magnifica Humanitas", calls for restraint in AI development.

He quotes J R R Tolkien from "The Return of the King."

“It is not our part to master all the tides of the world, but to do what is in us for the succour of those years wherein we are set, uprooting the evil in the fields that we know, so that those who live after may have clean earth to till. What weather they shall have is not ours to rule.”

Indeed.

Considering the effort to develop AI at breakneck speed, was Leo trying to draw a parallel between the Ring and an AI to rule them all?

 

[gleem]

Of all the major AI companies, Anthropic seems to be the most concerned about safety and the one doing the most to study alignment issues. It released the following document discussing its apprehensiveness using AI in improving AI. It remains concerned about the speed with which AI is being developed and supports an international pause in AI development.

https://www.anthropic.com/institute/recursive-self-improvement